Make a scanned PDF searchable: its words are recognised on your device. Free, with no account or uploads.
Drop a scanned PDF to make it searchable
Its words are recognised on your device and laid unseen over the pages, so you can search, select, copy and redact them. The pages look just as they did. You can also paste a file with ⌘V or Ctrl+V.
PDF
Stays on your device · up to 300 MB
Off, nothing is kept: closing the tab loses changes that were not downloaded. On, the file and your changes are kept in this browser until you close the file or throw the draft away. Nothing is sent anywhere.
What you can do
Scans you can searchThe words of scanned pages are recognised and laid unseen over them, each where it is seen. Search, select and copy them, or find and redact names and numbers.
The pages look the sameNothing is drawn: the picture of each page is left as it was. Pages that already have text are left alone unless you ask for them too.
Read back before downloadHow sure the reading was is said for every run, and the words laid are read back from the file before you download it. Undo takes them all away.
Nothing is uploadedThe words are read on your device, in any of 31 languages. Each is loaded from this site the first time it is used (a few megabytes) and kept for working offline.
This is the PDF editor, opened for recognising the words of scanned pages. Its other tools (text, pictures, shapes, highlights, redaction) are behind “All tools”.
A little help along the way
Follow the steps below. Every option is explained underneath.
37 s, no sound. On-screen captions describe each step; the same steps are written out here.
1Open the scanDrop the PDF, choose it from your device or paste it. A locked file asks for its password. The panel for recognising text is open from the start.
2Choose the pagesLeave Pages empty to read every page, or type pages such as 1-3, 7 or odd. Pages that already have text are left as they are, unless you turn on “Pages that have text too”.
3Recognise the textPress Recognise text. Each page is read in turn on your device, a few seconds a page, longer on a phone; Stop keeps the pages read so far. The panel then says how many words were found and how sure the reading was.
4Search and checkSearch the file or select its words: they lie where they are seen. Names and numbers that matter are worth a look, since a word can be misread.
5DownloadPress Download. The words laid are read back from the file first, and the check says so. The result is searchable in any PDF reader.
Every option explained
What recognising does
Each page is drawn as a picture and read by Tesseract, an open-source recogniser, running in your browser. The words it finds are written into the page in a font that draws nothing, each stretched to the width it has on the page. The picture is not changed, so the page looks exactly as it did, and nothing is uploaded.
How sure it was
Tesseract says how sure it is of every word. The panel gives the average and how many words it was unsure of. Clean, straight scans at 300 dots to the inch read best; handwriting, faint copies and photographs of pages at an angle read worse, and are read all the same.
Pages that have text
A page that already has text, from a recogniser that ran before or because it was never a scan, is left as it is, and the panel names it. Turn on “Pages that have text too” to read it anyway; its words are then in the file twice.
Turned pages
A page that is turned to be shown upright is read as it is shown, and its words run the way you see them.
Language
Choose the language the pages are written in: 31 are offered, among them Arabic, Chinese, Hindi, Japanese, Russian and Thai, as well as most languages of Europe. The language your browser says you read comes first. “Also written in” adds a second for pages that mix two; reading then takes longer. Arabic and Hebrew are read right to left, and Chinese, Japanese and Thai words are written together, as those languages write them. Text that runs down the page is not read.
Working offline
What recognises words is about 1 MB, and each language between 0.4 and 3.7 MB; the panel says how much the languages chosen take. It is loaded from this site the first time you recognise text in a language, not before, and then kept on this device, so it works offline afterwards.
Undo
A run is one step: one undo takes away every word it laid, and the page is as it was.
All tools
This page is the PDF editor, opened for recognising text. All tools shows the rest: text, pictures, shapes, highlights, and redaction, which can now find the words of the scan.